※ veridical.net
  • About
  • Publications
  • 3 Questions
  • Top N
  • Tags
  • Archives

Our Dexterous Future?

Yes, I’ve seen Humans Need Not Apply1 but am still not as pessimistic as Grey, even though large-language models have become a (very useful!) thing in the meantime. I’m of course open to changing my mind as the situation warrants it, but for now…so far, so good.

An artificial super-intelligence would be an obvious concern — I mean how could us simpletons keep it from enslaving us all (whether it be iRobot-style or Wall•E-style). But I’m not worried about super-intelligence (yet?), because the existing crop of AIs — LLMs in particular — aren’t intelligent at all, much less super, and I haven’t read anything about how they’d cross that chasm. But more on intelligence another time.

Another obvious concern (the one from the video) is robots taking all the physical jobs, which of course doesn’t require super-intelligence, but it would require ubiquitous utility. AI2 and Robots are replacing humans in some jobs, but they will also create/enable new jobs, and as long as there’s a class of tasks where humans have a comparative advantage this technological revolution seems likely to play out like all the rest (e.g., agriculture, printing, steam, computation): disruptive, society-shaping, likely productivity- and welfare-enhancing, and ultimately non-existential. (Meaning, I don’t dismiss that huge changes are coming; I just think we need the perspective that humanity has worked through challenges like this before, so we should should go forth with sanguinity, not despair).

That’s why this Construction Physics post about Robot Dexterity gave me even more optimism: it’s another category of work I hadn’t deeply considered that: (1) Robots suck at (the first is those that require intelligence), and (2) many humans genuinely enjoy (not these specific tasks of course). Here’s his list, which is delightful in its specificity and an opportunity to wonder at the capabilities of humans. (I mean think about it; can anything but a human do any of these? Perhaps a primate can peel an orange?)

  1. Put on a pair of latex gloves
  2. Tie two pieces of string together in a tight knot, then untie the knot
  3. Turn to a specific page in a book
  4. Pull a specific object, and only that object, out of a pocket in a pair of jeans
  5. Bait a fishhook with a worm
  6. Open a childproof medicine bottle, and pour out two (and only two) pills
  7. Make a peanut butter and jelly sandwich, starting with an unopened bag of bread and unopened jars of peanut butter and jelly
  8. Act as the dealer in a poker game (shuffling the cards, dealing them out, gathering them back up when the hand is over)
  9. Assemble a mechanical watch
  10. Peel a piece of scotch tape off of something
  11. Braid hair
  12. Roll a small lump of playdough into three smaller balls
  13. Peel an orange
  14. Score a point in ladder golf
  15. Replace an electrical outlet in a wall with a new one
  16. Do a cat’s cradle with a yo-yo
  17. Put together a plastic model, including removing the parts from the plastic runners
  18. Squeeze a small amount of toothpaste onto a toothbrush from a mostly empty tube
  19. Start a new roll of toilet paper without ripping the first sheet
  20. Open a ziplock bag of rice, pull out a single grain, then reseal the bag
  21. Put on a necklace with a clasp

An army of super-intelligent, dexterous robots would be concerning. I’m not holding my breath.


  1. Which is almost 11 years old! ↩

  2. Yes, I realize I just said, without proof, that what we call AI isn’t actually intelligent, but that’s the term in use, so that’s the term I use. ↩


  • « Straussian Reading

Published

May 26, 2025

by Adam Wuerl

Tags

  • AI 2
  • automation 4

Contact

Twitter
RSS
Mastodon
LinkedIn
Stack Exchange
Micro.blog
  • Powered by Pelican. Theme: Elegant by Talha Mansoor